WP1: Data acquisition, pre-processing and correction for topographic effects
Context
Assessing the rate and spatial pattern of forest cover change is challenging given the ruggedness and the inaccessibility of mountain areas. Remote sensing techniques are a privileged tool, even if they suffer from methodological challenges that have to be resolved by appropriate and advanced pre-processing techniques. Radiometric correction techniques are a necessary step for change detection analysis to obtain homogeneous time series of satellite data. This procedure should include sensor calibration, atmospheric and topographic corrections, and relative radiometric normalization (Vicente-Serrano et al., 2008). Topographic effects (Figure 1) have long been recognized as a problem for multispectral and multi-temporal vegetation classification in steep terrain as topography can bias the signal recorded by spaceborne optical sensors (Lu et al., 2008; Riaño et al., 2003; Richter et al., 2009a). Shadowing effects cause variations in land surface reflectance values due to the position of the sun. For the same land surface characteristics, slopes oriented away from (or in the direction of) the sun will appear darker (or brighter) compared to a planar surface (Richter and Schläpfer, 2011).
Figure 1: Different levels of topographic correction techniques in the Ecuadorian Andes (Landsat RGB 453).
Accurate forest cover monitoring in steep terrain is clearly compromised by terrain effects. When analysing time-series and/or multi-sensor images, apparent reflectance values can vary largely because of the position of the sensor, the sun zenith and azimuth angles that are all scene-specific. It may be difficult to distinguish forest cover changes from topographic bias (Huang et al., 2008). Although topographic disturbances have to be considered for an effective correction (Kobayashi and Sanga-Ngoie, 2008), a large majority of remote sensing studies that are based on the retrieval of reflectance values in 3D conditions are assuming flat surfaces.
Main results
1. Refinement of ATCOR3 topographic correction
A topographic correction of optical remote sensing data is necessary to improve the quality of quantitative forest cover change analyses in mountainous terrain. The implementation of semi-empirical correction methods requires the calibration of model parameters that are empirically defined. This study develops a method to improve the performance of topographic corrections for forest cover change detection in mountainous terrain through an iterative tuning method of model parameters based on a systematic evaluation of the performance of the correction. The latter was based on: (i) the general matching of reflectances between sunlit and shaded slopes (Figure 2) and (ii) the occurrence of abnormal reflectance values, qualified as statistical outliers, in very low illuminated areas. The method was tested on Landsat ETM+ data for rough (Ecuadorian Andes) and very rough mountainous terrain (Bhutan Himalayas) (Figure 2). Compared to a reference level (no topographic correction), the ATCOR3 semi-empirical correction method resulted in a considerable reduction of dissimilarities between reflectance values of forested sites in different topographic orientations (Figure 2). Our results indicate that optimal parameter combinations are depending on the site, sun elevation and azimuth and spectral conditions. We demonstrate that the results of relatively simple topographic correction methods can be greatly improved through a feedback loop between parameter tuning and evaluation of the performance of the correction model.
Figure 2: Left image: Comparison of level 1 (a and c) and 2 (b and d) false color composites (RGB 453) in the Ecuadorian and Bhutanese test sites where a decrease of the relief effect can be seen for the combination of atmospheric and topographic corrections in contrast with atmospheric correction only. Right image: Box and whiskers plots of level 1 (left) and level 2 (right) reflectances vs. aspect classes for band 4 and 7 for study sites of Ecuador (upper panels) and Bhutan (lower panels).
Additionnal information can be found in:
Balthazar, V., Vanacker, V., Lambin, E.F., 2012. Evaluation and parameterization of ATCOR3 topographic correction method for forest cover mapping in mountain areas. International Journal of Applied Earth Observation and Geoinformation, 18, 436-450.
2. Combination of atmospheric and topographic correction techniques
Mapping of vegetation in mountain areas based on remote sensing is obstructed by atmospheric and topographic distortions. A variety of atmospheric and topographic correction methods has been proposed to minimize atmospheric and topographic effects and should in principle lead to a better land cover classification. The purpose of this study was to evaluate the effect of coupled correction methods on land cover classification accuracy (Figure 3). Therefore, all combinations of three atmospheric (no atmospheric correction, dark object subtraction and correction based on transmittance functions) and five topographic corrections (no topographic correction, band ratioing, cosine correction, pixel-based Minnaert and pixel-based C-correction) were applied on two acquisitions (2009 and 2010) of a Landsat image in the Romanian Carpathian mountains. The accuracies of the fifteen resulting land cover maps (Figure 3) were evaluated statistically based on two validation sets: a random validation set and a validation subset containing pixels present in the difference area between the uncorrected classification and one of the fourteen corrected classifications. New insights into the differences in classification accuracy were obtained.
Figure 3: True color composite (RGB: bands 3, 2 and 1) and Maximum Likelihood classification of the 2009 image with a linear stretching: (a) no AC or TC with implementation of the GCPs; (b) TF with cosine correction and (c) TF with PBC correction.
First, results showed that all corrected images resulted in higher overall classification accuracies than the uncorrected images. The highest accuracy for the full validation set was achieved after combination of an atmospheric correction based on transmittance functions and a pixel-based Minnaert topographic correction. Secondly, class accuracies of especially the coniferous and mixed forest classes were enhanced after correction. There was only a minor improvement for the other land cover classes (broadleaved forest, bare soil, grass and water). This was explained by the position of different land cover types in the landscape. Finally, coupled correction methods showed most efficient on weakly illuminated slopes. After correction, accuracies in the low illumination zone (cos β ≤ 0.65) were improved more than in the moderate and high illumination zones. Considering all results, best overall classification results were achieved after combination of the transmittance function correction with pixel-based Minnaert or pixel-based C-topographic correction. Furthermore, results of this bi-temporal study indicated that the topographic component had a higher influence on classification accuracy than the atmospheric component and that it is worthwhile to invest in both atmospheric and topographic corrections in a multi-temporal study.
Additionnal information can be found in:
Vanonckelen S., Lhermitte S., Van Rompaey A., 2013. The effect of atmospheric and topographic correction methods on land cover classification accuracy, International Journal of Applied Earth Observation and Geoinformation, 24, 9-21.